Class-wise Thresholding for Robust Out-of-Distribution Detection.
Matteo GuarreraBaihong JinTung-Wei LinMaria A. ZuluagaYuxin ChenAlberto L. Sangiovanni-VincentelliPublished in: CVPR Workshops (2022)
Keyphrases
- landmark detection
- partial occlusion
- detection method
- post processing
- detection rate
- mild assumptions
- automatic detection
- detection algorithm
- spatial distribution
- anomaly detection
- threshold selection
- object detection
- probability distribution
- pairwise
- driver assistance systems
- reliable detection
- class membership
- noisy environments
- false positives
- detection accuracy
- data sets
- computationally efficient
- edge detection
- denoising
- image processing
- computer vision